Understanding AI Recognition Accuracy

Ginkgo’s AI can identify a wide range of household objects, but it’s not perfect. Understanding its strengths and limitations helps you get the best results.

What AI recognition does well

• Common household items: Tools, appliances, furniture, kitchenware, electronics
• Branded items: Products with visible brand names or logos are identified more accurately
• Items with distinctive shapes: A drill, a kettle, a lamp - things with recognisable forms
• Labelled items: Items with text, model numbers, or clear descriptions in frame

What it struggles with

• Generic or featureless objects: A plain cardboard box, a roll of tape, or a black cable are harder to identify
• Very similar items: Distinguishing between two similar-looking cables or tools can be difficult
• Poor photos: Blurry, dark, or partially obscured photos reduce accuracy significantly
• Very unusual items: Rare, custom, or specialised items may not be recognised

Improving recognition

If recognition is giving poor results for a category of item:

1. Improve the photo - more light, full item in frame, less background clutter
2. Include text in the shot - angle the photo to capture brand names or labels
3. Edit after recognition - accept the suggestion and edit the name to be more specific
4. Add manually - for items that are simply too generic to recognise, manual entry is quicker than multiple photo attempts

Feedback

If the AI frequently misidentifies a type of item you add often, let us know. We use feedback to prioritise improvements to the recognition model.

You can send feedback via Settings > Feedback or email support@getginkgo.uk.